UrielAPI Terms, Privacy, and Data Policy

Plain-language summary: UrielAPI does not train models on customer prompts, responses, files, API requests, or outputs. UrielAPI does not retain prompt or response content for model training. API content is processed transiently to provide inference and may be routed to upstream inference providers. UrielAPI retains limited operational metadata needed for billing, security, abuse prevention, debugging, reliability, and customer support.

1. Service

UrielAPI provides access to an OpenAI-compatible API gateway for AI model inference. UrielAPI provides commercially licensed model access through authorized provider relationships. Model availability, routing, context length, and pricing may vary by endpoint and provider configuration. The service is currently offered as a beta API service and may change without prior notice.

2. No Model Training on Customer Data

UrielAPI does not train, fine-tune, or improve models using customer prompts, completions, files, API request bodies, or generated outputs. Customer API content is not used to build or train any model owned or operated by UrielAPI.

3. API Content Handling

Prompt and response content is processed transiently for the purpose of generating the requested inference response. UrielAPI does not intentionally store full prompt or response content in its application database for normal API operation. Customers should not send sensitive personal, confidential, regulated, or legally restricted information unless they have independently determined that such processing is appropriate for their use case.

4. Operational Metadata

UrielAPI may store limited operational metadata, including customer account records, API key metadata, token counts, credit balance, request timestamps, selected provider, model name, request success or failure status, error summaries, latency, cost calculations, and similar usage records. This metadata is used for prepaid credit deduction, usage dashboards, abuse prevention, debugging, support, reliability analysis, and security.

5. Upstream Providers

Requests may be routed to upstream inference providers in order to deliver the service. Those providers may process API content according to their own infrastructure and policies. UrielAPI does not represent itself as OpenAI, ChatGPT, or an official OpenAI product.

6. Prepaid Credits and Billing

Access may be prepaid. Usage is deducted from customer credit based on model-specific input and output token usage, provider cost, and published pricing. Credits are not a bank account, security, deposit, or stored-value financial instrument. Used credits are generally non-refundable. Unused prepaid credit may be refunded on request, subject to abuse review, payment processor fees, fraud prevention, and already-completed usage.

7. Acceptable Use

Customers may not use the service for illegal activity, spam, malware, credential theft, exploitation, abuse, evasion of safety systems, harassment, or activity that violates applicable laws or third-party rights.

8. Availability

The service depends on upstream providers, routing infrastructure, and internet connectivity. UrielAPI does not guarantee uninterrupted availability, fixed latency, fixed model availability, or error-free operation during beta.

9. Retention and Deletion

Operational metadata may be retained as needed for billing, security, debugging, support, fraud prevention, tax/accounting records, and compliance. Customers may contact UrielAPI to request account review or deletion where applicable.

10. Limitation of Liability

The service is provided as-is. UrielAPI is not liable for indirect damages, lost profits, data loss, downtime, provider failures, output errors, or business interruption.